Start here
You belong in this room if…
You do not need a CS degree or a model cheat sheet. You need curiosity about how work actually gets done when software starts acting on your behalf.
- You keep hearing “agents” and want a grounded definition before you buy or build
- Your team uses Copilot-style chat but wonders what “agent mode” really changes
- You supervise projects and need vocabulary security and finance will trust
- You are an engineer who wants the business side in the same conversation
A chatbot answers a question. An agent takes a goal, plans steps, uses tools, and reports back. Once you feel that difference on your own machine, the rest of the industry noise gets quiet.
2hLive on Zoom · Thursday evenings (PST)
3Modules with a workshop each
Per dateSeat price & minimum seats set per Thursday
Schedule
Pick a Thursday cohort
Loading upcoming dates…
Pricing
- Select a session to see that date's seat price and minimum seats.
The map
Three ideas that unlock every agent conversation
Skip the buzzword bingo. Every serious deployment comes down to permission (what it may touch), context (what it can see), and tokens (what you pay to keep the window full). We teach them as lived habits, not slide fodder.
-
Agents
Permission to act
Plans, runs commands, reads output, repeats — with boundaries you can explain to your boss and your security team.
-
Context
What the model can see
Files, rules, tool output — not your screen, not “what you meant.” Learn to write requests that survive the first step.
-
Tokens
What you pay to load
Context windows refill every step. Spend on decisions and diffs, not repeated clutter.
Not this
- Prompt cheat sheets
- “Top 10 ChatGPT tricks”
- Model leaderboard debates
This course
- Systems thinking you can operationalize
- Live agent loops on your machine
- Language that matches how production harnesses are built
Syllabus
One night, three sections
Each Thursday is a single 2-hour Zoom that follows the same arc as our presentation deck: teach → workshop for Agents, Context, and Tokens. You leave with notes, not slides you will never open again.
Input, steps, output, cost — the loop behind every agent. Why typing commands beats clicking through UIs for auditable work.
- Grant system access safely (scope, not “full admin”)
- Plan → act → read → repeat
- Try it in Cursor, Claude Code, or similar
Workshop: Run one boring weekly task as an agent loop and name what you would log for compliance.
Context is everything in the window: your ask, the rules, the files you attach, command output. What never enters is not “hidden” — it simply does not exist for the agent.
- Vague vs precise requests on the same task
- Context as a desk that refills every step
- When to start a fresh chat vs continue
Workshop: Same job, two briefs — measure which one finishes with fewer loops.
How models read text, why inference is “next token,” and how that maps to dollars and latency. Learn this so you can optimize context and execution — what to load, what to skip, what each loop costs.
- One task per chat, one role per agent
- Tool output hygiene (diffs over dumps)
- Estimate cost for a real workflow
Workshop: Ballpark token spend for a workflow you already run manually.
Who it’s for
Built for operators, not only engineers
- Leaders who need to buy or supervise agent work responsibly
- Consultants and PMs translating “AI” into scoped projects
- Engineers who want a shared vocabulary with the business
- Teams evaluating Cursor-style agents vs chat-only copilots
You leave with
Outcomes
- A checklist for safe agent scope on a real host
- Before/after examples of context that actually ships work
- A token budget habit for one recurring task
- Language to align with our AI Report findings or your internal rollout
Your guide
Matt Potter
Speaker, developer, and AI engineer. Matt designs production harnesses for Agent Accelerated and teaches the same permission · context · purpose stack he uses on stage and in client work. Expect demos, not decks.
Read the full bio
Format
What a session feels like
- Story — a real task, told without jargon
- Demo — watch an agent loop on a laptop
- Workshop — you run the loop on a boring job from your week
- Debrief — what to log, what to scope, what to budget
Delivered live on Zoom (link after confirmation). Bring a laptop and one mundane task (status email, file shuffle, ticket triage, expense report).
Questions before you commit? Contact us or RSVP below and note your team size in the form.
RSVP
Reserve for How Agents Work on Zoom. Choose a Thursday below; we email payment and join details once your cohort hits the minimum.